Abstract

This Project investigates COVID-19 data through a rigorous statistical lens, aiming to uncover patterns, influences, and trends in the pandemic’s trajectory. The study meticulously examines various visualizations and statistical methods to elucidate critical insights. It explores the impact of WHO regions on reported new cases across countries and analyzes temporal trends in new deaths. Leveraging a dataset spanning the pandemic’s timeline, the research employs a meticulous data preparation process, acknowledging biases and ensuring robustness. A suite of compelling visualizations, including line plots, mosaic plots, map visualizations, 3D scatter plots, and a correlogram, complements the analyses, providing a comprehensive understanding of COVID-19 data dynamics. The paper applies hypothesis tests, aligning with research questions and statistical methodologies, yielding nuanced insights. The results and conclusions drawn from the analyses offer a clear and substantiated perspective on the pandemic’s patterns, facilitating informed understanding and potential future research avenues.

# Load CSV data into a variable (change file path if necessary)
covid_data <- read.csv("WHO-COVID-19-global-data.csv", stringsAsFactors = FALSE)

Understanding the data:

During the preliminary data exploration phase, a pivotal step in comprehending the COVID-19 data set involved crafting a set of diverse visualizations. These visual representations served as essential tools to unravel underlying trends, intricate patterns, and interrelationships embedded within the data set.

# Check the structure of the loaded data
str(covid_data)
## 'data.frame':    341754 obs. of  8 variables:
##  $ Date_reported    : chr  "2020-01-03" "2020-01-04" "2020-01-05" "2020-01-06" ...
##  $ Country_code     : chr  "AF" "AF" "AF" "AF" ...
##  $ Country          : chr  "Afghanistan" "Afghanistan" "Afghanistan" "Afghanistan" ...
##  $ WHO_region       : chr  "EMRO" "EMRO" "EMRO" "EMRO" ...
##  $ New_cases        : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ Cumulative_cases : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ New_deaths       : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ Cumulative_deaths: int  0 0 0 0 0 0 0 0 0 0 ...

Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.

# Check for missing values
summary(covid_data)
##  Date_reported      Country_code         Country           WHO_region       
##  Length:341754      Length:341754      Length:341754      Length:341754     
##  Class :character   Class :character   Class :character   Class :character  
##  Mode  :character   Mode  :character   Mode  :character   Mode  :character  
##                                                                             
##                                                                             
##                                                                             
##    New_cases       Cumulative_cases      New_deaths       Cumulative_deaths
##  Min.   : -65079   Min.   :        0   Min.   :-3520.00   Min.   :      0  
##  1st Qu.:      0   1st Qu.:     2994   1st Qu.:    0.00   1st Qu.:     22  
##  Median :      0   Median :    38563   Median :    0.00   Median :    415  
##  Mean   :   2260   Mean   :  1569014   Mean   :   20.45   Mean   :  18669  
##  3rd Qu.:    132   3rd Qu.:   461936   3rd Qu.:    1.00   3rd Qu.:   6116  
##  Max.   :6966046   Max.   :103436829   Max.   :44047.00   Max.   :1144877
library(ggplot2)
## Warning: package 'ggplot2' was built under R version 4.3.2
library(plotly)
## Warning: package 'plotly' was built under R version 4.3.2
## 
## Attaching package: 'plotly'
## The following object is masked from 'package:ggplot2':
## 
##     last_plot
## The following object is masked from 'package:stats':
## 
##     filter
## The following object is masked from 'package:graphics':
## 
##     layout
library(reshape2)
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
# Line plot for cumulative cases over time
ggplot(covid_data, aes(x = Date_reported, y = Cumulative_cases, group = 1)) +
  geom_line(color = "darkslateblue") +
  labs(title = "Cumulative COVID-19 Cases Over Time",
       x = "Date",
       y = "Cumulative Cases") +
  theme_minimal()

# Mosaic plot for region-wise distribution of new cases
ggplot(data = covid_data, aes(weight = New_cases, x = "", fill = WHO_region)) +
  geom_bar(width = 1) +
  coord_polar("y", start = 0) +
  labs(title = "Region-wise Distribution of New Cases",
       fill = "WHO Region") +
  theme_void() +
  theme(legend.position = "bottom")

library(plotly)
library(dplyr)

# Assuming covid_data is your data frame
plot <- plot_ly(data = covid_data, x = ~New_cases, y = ~Cumulative_cases, z = ~New_deaths, 
                type = "scatter3d", mode = "markers", marker = list(size = 5, color = "darkslateblue"))

plot <- plot %>% layout(scene = list(xaxis = list(title = "New Cases"),
                                     yaxis = list(title = "Cumulative Cases"),
                                     zaxis = list(title = "New Deaths")),
                        title = "3D Scatter Plot of COVID-19 Data")

plot  # Display the plot
library(corrplot)
## corrplot 0.92 loaded
# Correlogram for correlations between variables
cor_matrix <- cor(covid_data[, c("New_cases", "Cumulative_cases", "New_deaths", "Cumulative_deaths")])
corrplot(cor_matrix, method = "circle", type = "lower", tl.col = "black")

library(GGally)
## Warning: package 'GGally' was built under R version 4.3.2
## Registered S3 method overwritten by 'GGally':
##   method from   
##   +.gg   ggplot2
# Advanced Visualization with GGally for pairwise scatter plots
ggpairs(covid_data[, c("New_cases", "Cumulative_cases", "New_deaths", "Cumulative_deaths")])

Hypothesis 1:

Purpose: The number of new cases reported in a country is influenced by the WHO region.

Null Hypothesis (H0): The WHO region does not have any significant influence on the number of new cases reported in a country.

Alternative Hypothesis (H1): The WHO region has a significant influence on the number of new cases reported in a country.

# Fit a linear model (ANOVA) to test the influence of WHO region on New cases
model <- lm(New_cases ~ WHO_region, data = covid_data)

# Perform ANOVA to test for differences in new cases among WHO regions
anova_result <- anova(model)

anova_result

Interpretation: The p-value obtained (p < 0.001) suggests strong evidence against the null hypothesis that the WHO region does not have a significant influence on the number of new cases reported in a country. Therefore, based on this analysis, it appears that the WHO region has a statistically significant impact on the reported new cases.

Hypothesis 2:

Purpose: There is a temporal trend in the number of new deaths reported over time.

Null Hypothesis (H0): There is no temporal trend or significant change in the number of new deaths reported over time.

Alternative Hypothesis (H1): There exists a temporal trend or significant change in the number of new deaths reported over time.

# Assuming your data frame is named 'covid_data'
# Convert Date_reported column to Date format if it's not already in Date format
covid_data$Date_reported <- as.Date(covid_data$Date_reported)

# Fit a linear regression model to assess the trend in New deaths over time
model <- lm(New_deaths ~ Date_reported, data = covid_data)

summary(model)
## 
## Call:
## lm(formula = New_deaths ~ Date_reported, data = covid_data)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
##  -3530    -26    -16     -8  44036 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)    4.497e+02  1.273e+01   35.33   <2e-16 ***
## Date_reported -2.261e-02  6.703e-04  -33.73   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 163.1 on 341752 degrees of freedom
## Multiple R-squared:  0.003318,   Adjusted R-squared:  0.003315 
## F-statistic:  1138 on 1 and 341752 DF,  p-value: < 2.2e-16

Interpretation: The linear regression analysis shows a statistically significant but very weak negative relationship between the date reported and the number of new deaths. However, the model explains only a small fraction of the variance in new deaths based on the date reported, suggesting that other factors not included in this model might influence the number of new deaths reported over time.

Hypothesis 3:

Purpose: To conduct hypothesis tests without assuming a specific distribution.

Null Hypothesis (H0): There is no difference in the average number of new cases reported between the first half and the second half of the data set.

Alternative Hypothesis (H1): There is a significant difference in the average number of new cases reported between the first half and the second half of the dataset.

# Calculate the midpoint of the dataset
midpoint <- nrow(covid_data) / 2

# Extract new cases for the first and second halves
first_half <- covid_data[1:midpoint, "New_cases"]
second_half <- covid_data[(midpoint + 1):nrow(covid_data), "New_cases"]

# Calculate the observed difference in means
observed_diff <- mean(first_half) - mean(second_half)

# Function to perform permutation test
permutation_test <- function(iterations = 1000) {
  diffs <- numeric(iterations)
  for (i in 1:iterations) {
    combined <- sample(c(first_half, second_half))
    perm_first_half <- combined[1:length(first_half)]
    perm_second_half <- combined[(length(first_half) + 1):length(combined)]
    diffs[i] <- mean(perm_first_half) - mean(perm_second_half)
  }
  return(diffs)
}

# Perform permutation test
set.seed(123)  # For reproducibility
perm_diffs <- permutation_test(iterations = 1000)

# Calculate p-value
p_value <- mean(abs(perm_diffs) >= abs(observed_diff))

p_value
## [1] 0

Interpretation: A p-value of 0 suggests that among the randomly permuted differences in means generated by the permutation test, none were as extreme as the observed difference. In other words, the observed difference in the average number of new cases between the first and second halves of the dataset is so far from what would be expected under the null hypothesis (where there’s no difference) that it essentially indicates strong evidence against the null hypothesis.

Therefore, we can reject the null hypothesis and conclude that there is a statistically significant difference in the average number of new cases between the two time periods (first and second halves of the data set).